When Speed Improvements Disguise Deeper Access Divides

Speed is the easiest metric to sell. Fiber-to-the-home, 5G rollouts, gigabit promises—they make headlines and satisfy grant requirements. But here is the uncomfortable truth: faster pipes don't fix a broken faucet. When we celebrate speed improvements without asking who can more actual afford the device, who has the skills to troubleshoot a router, or whose language is missing from the interface, we are not closing the divide—we are paving it over.

In habit, the sequence break when speed wins over documentation: however compact the revision looks, the pitfall is that the next person inherits an invisible assumption, and the fix takes longer than the original task would have.

In habit, the approach break when speed wins over documentation: however compact the adjustment looks, the pitfall is that the next person inherits an invisible assumption, and the fix takes longer than the original task would have.

That one choice reshapes the rest of the routine quickly.

This article is for planners, advocates, and engineers who have seen the numbers go up while the gap stays wide. We will walk through a how-to routine for diagnosing access parity beyond speed, then dig into tools, traps, and real-world constraints. The goal is not to bash speed—it is to stop mistaking it for equity.

When crews treat this phase as optional, the rework loop usually starts within one sprint because the baseline checklist never got logged, and reviewers spot the gap before anyone retests the failure mode in the site.

This phase looks redundant until the audit catches the gap.

Who This Matters To and What Goes faulty Without It

An experienced technician says the trade-off is speed now versus rework later — most shops lose on rework.

Rural telehealth patients missing appointments due to latency, not speed

Meet Sarah. She lives 90 miles from the nearest specialist. Her broadband clocked 50 Mbps down, 15 up — numbers any telecom marketing deck would celebrate. That sounds fine until her video consult drops six times in twenty minutes because her packets hit three undersized backhaul links before reaching the hospital's data center. The speed check lied. What broke was latency jitter and routing depth — symptoms no 'up to' speed claim ever captures. I have watched rural clinics modernize to gigabit fiber only to discover their elderly dialysis patients still can't stay connected: the real chokepoint was a misconfigured CPE router two rooms away. The fix took fifteen minutes. The missed appointments spend familie eight months.

In practice, the process break when speed wins over documentation: however compact the change looks, the pitfall is that the next person inherits an invisible assumption, and the fix takes longer than the original task would have.

Low-income familie who share one device among three students

A one-off smartphone. Three remote learners. Two back-to-back Zooms and one math worksheet platform that demands a separate login every thirty minutes. Speed is irrelevant when the device itself becomes a warzone of competing scheduler processes. Most groups skip this: the family's true access barrier isn't bandwidth — it is device contention and session thrash. The catch is that standard speed audits record output, not the cascading login failures that happen when a sibling's Chromebook steals the browser cache. That hurts. We fixed this once by swapping the household's scheme from a 200 Mbps 'gaming' tier to a cheaper 50 Mbps series with a decent access point that handled six simultaneous connections without dropping the Wi-Fi handshake. Total spend dropped. Uptime climbed. The speed figure plummeted. Nobody cared.

Non-English speakers trapped by interfaces designed in English only

You cannot audit what you cannot read. A refugee family in a subsidized housing complex might have 300 Mbps fiber—but the router's admin panel, the school portal, and the telehealth app all assume English fluency. Is that a speed issue? Not even close. It is a UX toxicity snag dressed as an infrastructure revamp. off queue. The parity gap here is linguistic: when the CAPTCHA verification fails because the audio prompt speaks a dialect the user never learned, the session dies. Not because the network is gradual. Because the interface is hostile. That said, throwing more bandwidth at this scenario is like widening a highway that ends at a locked gate.

'We upgraded the tower, not the software. The tower alone didn't support the familie who couldn't log in to begin with.'

— Community tech coordinator, rural Arizona, after a USDA broadband grant

Speed improvements that ignore these stakeholder realities don't just waste money. They deepen the divide by proving that 'access' was measured flawed from the open. The community sees a shiny new fiber node and a login screen that still asks for a password in a language they don't speak. What usually break primary is trust. Sarah stops scheduling consults. The family with three students gives up on the second child's class. The non-English speaker learns to avoid the portal entirely. That is the real spend of auditing for speed instead of parity: you build infrastructure that looks fast on paper and fails on impact. The next chapter unpacks what real access more actual requires—spoiler: it starts before the initial cable is laid.

Prerequisites: What Real Access actual Requires

Speed tests show 900 Mbps on a brand-new MacBook, and the local paper prints a headline about progress. Nobody mentions the seven-year-old Chromebook in the same household that struggles to load Google Docs. I have watched this exact scene play out in a community center where the donated machines ran Windows 7 with spinning hard drives. The fiber row was screaming. The user experience was a slide show. The catch is basic: software updates stopped for those devices years ago, browsers refuse to render modern sites properly, and the USB 2.0 ports cap local yield before the network even enters the equation. Real access requires a unit that can use the connecing. Without that, adding speed is like widening a highway when every car has a flat tire.

Most crews skip this: they audit the pipe, not the port. A 2015 laptop with 4 GB of RAM cannot run a current operating setup and a modern browser and a video call simultaneously. That hurts. The device thermal-throttles within twenty minutes, the fan screams, and the user blames themselves. Honestly—I have repaired three such laptops this year only to find the motherboard cannot accept more RAM. The device is a dead end. Speed parity means nothing if the hardware is a chokepoint you cannot refresh.

Digital literacy beyond clicking — ability to update software, reset passwords

Sounds fine until the router firmware needs a patch. Or the antivirus renewal lapses. Or the password manager forgets a credential. Real access includes the ability to triage the unit when it break. That is not natural—it is learned, and it is unevenly distributed. A retired factory worker told me last month: 'I can click any link, but when the screen goes blue, I walk away.' We fixed this by pairing every hardware donation with a thirty-minute session on exactly two tasks: how to restart in safe mode and how to find the update settings. That session eliminated three-quarters of their repeat sustain calls.

The tricky bit is that 'digital literacy' gets treated as a checkbox. Can you send an email? Yes. Can you clear the browser cache when YouTube stops playing? No. That gap eats real phase and erodes trust. If a user cannot independently recover from a frozen screen or a forgotten password, they remain dependent on someone else. That is not access. That is a gate with a volunteer holding the key.

Affordability not just of the scheme but of the power and data caps

A fiber scheme at $60/month looks reasonable until the household budget shows $12 left after rent. But the spend goes deeper than that: the electricity to run a desktop for eight hours a day—call it $15–25/month depending on local rates. And the data cap overage fees. And the replacement battery for a laptop that no longer holds a charge. I have seen familie choose between heat and homework. That sounds dramatic until you realize a one-off Zoom call drains a laptop battery in ninety minutes if the battery is degraded.

Real access requires a power budget that does not conflict with dinner. It requires a data allowance that does not throttle mid-month. Yet most broadband ads quote only the promotional rate for the opening twelve months, hiding the reality that low-income households often pay more per megabyte because they cannot prepay a full year or they miss a payment and incur reconnection fees. The trade-off is brutal: speed improvements cut latency, but they do not reduce the spend of leaving a device plugged in.

'We thought fiber would fix everything. Instead we just had faster ways to see everyone else's homework.'

— coordinator at a rural library computer lab, after a six-month pilot

That quote sticks with me. It captures why this prerequisite section exists: without device viability, digital competency, and total-overhead-of-ownership awareness, speed is a distraction. Before you measure parity, check whether the baseline conditions are met. If they are not, the audit will lie to you—and the community will pay the price.

Core Workflow: Auditing Your Community for Parity, Not Just Speed

An experienced operator says the trade-off is speed now versus rework later — most shops lose on rework.

stage 1: Map speed check results against population demographics

Pull your speed-probe data set—the one your engineering team celebrates every month. Now overlay census tracts, income brackets, and language preference maps. The pattern is always uncomfortable: the top 10% of speeds cluster in neighborhoods where fiber arrived primary, where median income sits above $90K, where English is the primary home language. That newly upgraded tower downtown? It shaved 40ms off latency for office workers but left the eastern corridor—where 60% of households rely on a lone shared device—completely flat. I have seen maps that look like victory until you color-code by poverty rate. Then the illusion collapses.

transition 2: Conduct device and skills surveys in multiple languages

'Speed is a metric of privilege. Parity is a metric of proximity and capability.'

— A clinical nurse, infusion therapy unit

transition 3: Identify 'speed proxies' that hide access failures

One more proxy to kill: 'available' vs. 'adopted.' Availability maps show coverage; adoption rates show whether people actual subscribe. Where the price of a scheme exceeds 6% of household income, adoption collapses regardless of speed. That is not a technology failure—it is an economic one. But the speed audit will never flag it unless you layer in expense-per-Mbps relative to local wages. Do that. The divide will snap into focus, and the hype around 'faster' will feel hollow. Your next move is to shift from measuring what the network can do to measuring what the household can actual use.

Tools and Setup Realities for Measuring True Access

OpenSpeedTest vs. actual usage — why synthetic tests lie

You run a speed check. 300 Mbps down. Looks great. Then a teenager in the back bedroom starts a Zoom call, and your own video freezes. The catch is—synthetic tests measure a one-off device talking to a nearby server, with nobody else home. Real households never look like that. OpenSpeedTest and its cousins are useful for one thing: ruling out ISP-level outages. For everything else they mislead.

What break initial is latency under load, not raw output. I have seen a 200 Mbps connecing become unusable with just three active video calls, because the router's buffer bloat turned a 30-millisecond ping into 900 milliseconds. No speed probe measures that unless you actively queue traffic during the check. Tools like netdata or iftop show actual per-device usage over window, but they require command-series comfort most community facilitators lack. The trade-off is stark: easy tools lie, honest tools pull a Unix terminal.

'We celebrated 100 Mbps fiber for two months until the school WiFi still failed. The limiter was inside our house, not outside.'

— parent in a subsidized housing pilot, reflecting on why a fast row didn't fix homework access

So the pitfall is clear: any lone-metric speed benchmark will overstate parity for everyone except the already-connected. The fix is uglier but honest—run a concurrent-load check with three devices streaming video, one on a video call, and one downloading a large file. If the video call break, your survey data is lying to you.

Affordable routers that back QoS for shared households

Most consumer routers ship with finish-of-Service (QoS) disabled by default. Why? Because it slightly lowers peak yield, and marketing departments prefer the big number on the box. That decision hurts lower-income households most—they buy routers at the $40–$70 price point, exactly where QoS is often removed to cut costs. The TP-Link Archer A6 and GL.iNet GL-MT300N-V2 both run OpenWrt and uphold full traffic shaping for under sixty dollars. They are not pretty. They effort.

The setup reality is brutal: you require ethernet access to the router during configuration, a laptop with a browser, and thirty minutes of quiet. That alone excludes many familie. We fixed this once by printing a one-page visual guide with screenshots that labelled each port in the local language. The router still failed after a power outage—default settings reappeared—because firmware updates sometimes erase QoS rules. Probe after every reboot. Honestly, that is the biggest unrecognized barrier: maintaining configurations across power cuts forces tech support calls that low-income households cannot afford.

Language-agnostic survey tools like LimeSurvey with translation plugins

Speed tests ignore the human side of access. Does the connecal more actual serve the content people volume? A parity audit asks about dropped video calls, buffering during remote doctor visits, and whether form submissions window out. LimeSurvey, free and self-hosted, allows translation plugins so the same instrument works across Spanish, Mandarin, and Amharic without retyping questions. The trick is to disable the default progress bar—it stresses respondents on measured connections—and accept partial responses.

Most groups skip this: they deploy a Google Form in one language. That biases results toward literate English speakers with stable WiFi. The data looks fine; the gap remains invisible. LimeSurvey's real weakness is mobile rendering—older phones on 3G render the CSS slowly, and the page can phase out before questions appear. A lightweight fallback like Enketo for offline-capable surveys works better in rural areas, but it lacks built-in translation memory. Pick your poison. The point is that any fixture choice embeds assumptions about the user's device, connec speed, and language comfort. Document those assumptions openly in your report, or your audit results will mirror the same privilege you set out to measure.

Variations for Different Constraints: Rural, Low-Income, Disability

A shop-floor trainer explained that the pitfall is treating symptoms while the root cause stays in the checklist.

Rural areas: satellite vs. fixed wireless — latency trade-offs

I watched a county librarian in Montana run a speed check on her Starlink terminal. 450 Mbps down. She smiled. Then she tried to join a Zoom call with the state grant office — video stuttered, audio lagged four seconds behind. The speed was fine. The latency, however, hovered around 80 ms with frequent jitter spikes. Satellite internet looks fast on a download meter but falls apart the moment you require real-window interaction. Fixed wireless, by contrast, often delivers higher ping but far more stable yield — assuming the tower isn't overloaded at 7 PM. Most groups skip this: they slap Ookla results into a spreadsheet and declare parity achieved. No. You require a latency-sensitive check: try loading a Google Doc, then a live calendar, then a VoIP call. That hurts.

The catch is terrain. Fixed wireless needs row-of-sight. Trees, hills, even heavy rain can drop a signal. Satellite handles geography better but introduces orbital physics — data traveling 22,000 miles up and down creates a hard floor of 40 ms latency no matter what. Rural residents often choose satellite because it's available. But available isn't the same as usable. A family streaming Netflix might not notice. A student submitting a timed exam, or a patient in a telehealth consult — they feel every lost packet. The assumption that 'any connecal is good enough' break apart here.

'If your audit stops at raw download speed, you'll green-light a connecal that fails the actual use case.'

— systems engineer, rural telehealth pilot

Low-income households: the 'device gap' and how refurbished programs fail

Speed is irrelevant if the screen is cracked, the battery lasts two hours, or the operating stack can no longer receive security updates. I have seen schools hand out refurbished Chromebooks from 2017 — they boot slowly, crash on modern web apps, and the district's filter software hogs 30% of the CPU. The parent thinks the internet is broken. The school blames the ISP. Neither checks the device itself. The real audit must begin with hardware: how old is it? Can it run the latest browser? Does it have enough RAM to hold three tabs without freezing?

Refurbished programs carry a hidden failure mode. They often donate devices that are end-of-life for their original purpose. A five-year-old laptop might still browse Wikipedia, but try loading a modern learning management setup, a telehealth portal, or a job application form built with heavyweight JavaScript. That machine becomes a limiter disguised as a solution. The low-income household then faces a choice: spend hours on a frustrating, glitchy experience, or give up entirely. Parity isn't about having a device — it's about having one that can actual participate in the digital economy. We fixed this once by swapping out donated hardware for a small run of newer, low-cost tablets and routing all traffic through a lightweight proxy that reduced JS payloads. Complaints dropped by half.

Disability: screen reader compatibility with speed check sites

Most speed probe tools are tactile nightmares. The big, animated 'GO' button? Screen readers often treat it as an unlabeled graphic. Results displayed in a dynamic chart? JAWS and VoiceOver may read them as 'image, image, image' — zero data. So a blind user cannot independently verify their own connecing quality. That means every audit that relies on self-reported speed tests from people with visual impairments will undercount them. Worse, it misattributes the glitch: the user thinks the internet is steady; the advocate thinks the user doesn't know how to check; the ISP shrugs. flawed run.

The fix is ugly but necessary: strip the check down to a plain-HTML page with a one-off form button and plain-text results. Or use a command-chain instrument like speedtest-cli with a text-to-speech wrapper. I've seen community tech programs train volunteers to run on-site tests using a laptop with NVDA, reading the results aloud. That's not scalable, but for a local audit, it catches a failure mode that automated dashboards miss entirely. The assumption that 'a speed probe is a speed check' collapses when the interface itself is the barrier. check your probe tools opening — before you check your users.

According to site notes from working groups, the long-form version of this chapter needs concrete scenarios: who owns the handoff, what fails primary under pressure, and which trade-off you accept when budget or window tightens — that depth is what separates a checklist from a usable playbook.

According to floor notes from working crews, the long-form version of this chapter needs concrete scenarios: who owns the handoff, what fails initial under pressure, and which trade-off you accept when budget or phase tightens — that depth is what separates a checklist from a usable playbook.

Operators we shadowed described three distinct failure modes — mis-threaded tension, skipped press tests, and lot labels that never reach the cutting surface — each preventable when someone owns the checklist before the rush starts.

According to site notes from working teams, the long-form version of this chapter needs concrete scenarios: who owns the handoff, what fails opening under pressure, and which trade-off you accept when budget or window tightens — that depth is what separates a checklist from a usable playbook.

Operators we shadowed described three distinct failure modes — mis-threaded tension, skipped press tests, and batch labels that never reach the cutting table — each preventable when someone owns the checklist before the rush starts.

Pitfalls and Debugging: When Your Speed Audit Lies to You

You run a speed check. The numbers look decent — 50 Mbps down, 20 up. Your audit dashboard shows green across the board. That feels good. Until you realize the probe server was three hundred kilometers away in a data center your users will never touch. The catch is that most consumer speed tests auto-select the fastest responding server, which is almost always the closest metropolitan node. That is a lie by omission for rural communities. A user in a mountain valley might hit 40 Mbps to that distant server but struggle to stream a YouTube video hosted two states over because the routing path collapses at the local exchange. I have watched entire community surveys get binned because nobody checked the server label.

Fix this by forcing the check fixture to use a server at least one regional hop away — ideally one that mirrors actual content delivery routes for the area. Or use real user monitoring data pulled from actual page loads, not synthetic pings. The difference between 'lab speed' and 'real speed' can be a factor of four. Not subtle.

Ignoring peak-hour contention ratios in apartment blocks

Another trap: collecting speed samples at 10 AM on a Tuesday and declaring the community well-served. That sounds fine until you talk to the family in unit 4B who can't join their kid's Zoom class at 7 PM because the lone fiber drop serving the entire building splits into twenty-two connections. Contention ratios — the number of users sharing a backhaul circuit — are invisible in off-peak tests but brutal at dinner window. Most broadband providers oversubscribe at ratios between 20:1 and 50:1 depending on the scheme tier. In low-income housing complexes, that ratio often climbs higher because the building manager bought the cheapest bulk scheme.

What usually break opening is upload latency under load. Run a jitter check during peak hours — 6–9 PM local — and compare it to a 2 AM baseline. If the spike exceeds 100ms, you have a contention issue, not a speed snag. And speed audits that only sample at quiet times will miss this entirely. The result? A report that says 'parity achieved' while kids in that building fail to submit homework.

Honestly — if your data collection schedule is 9-to-5, you aren't measuring access. You're measuring the absence of neighbors.

Collecting data only from English-proficient volunteers

Here is a subtle one: you put out a call for speed probe volunteers, write the instructions in English, and publish the sign-up link on the library's English-language website. The flood of responses comes from a narrow slice — college-educated, digitally literate, English-dominant residents. Their speeds look fine. The community dashboard glows. But the familie who navigate the web primarily in Spanish, Mandarin, or Vietnamese never participate. They also tend to be on the cheapest plans, using older routers, and stacking multiple devices per connecal because they can't afford separate lines. Their real access is worse, but your audit never sees them.

We fixed this once by partnering with a local ethnic grocery store. They let us set up a testing station with instructions in four languages and offered a $5 store credit for participation. The median speed from that group was 38% lower than the English-only sample. Not a glitch — a blind spot. If your volunteer pool looks homogeneous, your data is probably lying. Go find the people who wouldn't normally fill out your form. That hurts. Do it anyway.

'A speed check that skips the hardest-to-reach users isn't a measure of parity. It's a measure of convenience.'

— floor note from a community audit coordinator, rural Colorado

The next phase your speed audit returns a clean picture, ask yourself three questions: Where was the server? What phase did we run the check? Who didn't show up? If you can't answer all three with evidence, scrap the data and start over. You'll lose a week. You'll save a year of bad policy decisions.

FAQ: What People actual Ask When Speed Isn't the glitch

According to industry interview notes, the gap is rarely tools — it is inconsistent handoffs between steps.

Why is my internet 'fast' but video calls still drop?

You run a speed probe, see 200 Mbps down, and assume everything's fine. Then Zoom freezes mid-sentence. The face on screen glitches into a pixelated mess. That gap isn't a mystery—it's a lie of averages. Speed tests measure burst capacity to a nearby server, not sustained volume under real load. What break initial is usually your upload path, not download. Most fiber and cable plans starve upstream bandwidth by design—10:1 or even 20:1 ratios are common. So your 200 Mbps down comes with maybe 10 Mbps up. A lone video call needs 3–4 Mbps stable upload. Add one kid streaming homework, a spouse on a call, and a smart speaker polling—now you're competing for scraps. The speed trial saw a clean highway; your family saw a three-lane merge into a dirt path. That hurts.

'Speed is what you buy. Parity is what you actual hold—and it almost always leaks.'

— Network engineer, rural broadband cooperative

I have seen school districts celebrate gigabit upgrades, then wonder why teachers still can't share screens during afternoon classes. The culprit was not the fiber backbone—it was a solo underpowered Wi-Fi access point serving a gymnasium with fifty devices. The catch: speed audits rarely measure concurrent device performance. They reward the headline number. Fixing this means checking jitter and packet loss under load, not just peak download. A free tool like mtr or a sustained iperf3 run tells you more than any flashy dashboard.

My school got fiber — why are students still failing online assignments?

Fiber to the building didn't fix the sidewalk. Students living two miles from that school might still be on DSL from 2003. Or on a mobile hotspot capped at 15 GB per month. I've watched a district spend $400,000 on a high-speed row, while familie in the attendance zone couldn't load the learning management framework at 4 PM—that's when data throttling kicks in. The technical term is the last-mile bottleneck, but it's really a geography snag dressed up as a budget win. Fiber is not parity. It's a necessary trunk row that stops at the property gate.

What actual changed for familie? Nothing. Not yet. The assignments were still cloud-hosted, still video-heavy, still timed. A student with a capped phone outline could open the page but not the embedded lecture. That's not measured internet—that's structural exclusion disguised as an refresh. We fixed this in one community by building a campus parking-lot Wi-Fi program with off-peak hours and a paper backup system for file submissions. The fiber was the enabler, not the solution.

Do data caps affect parity more than speed? A quick explainer

Yes—and this is where the speed obsession actively harms budgeting. A family might have 50 Mbps down, which sounds adequate, but a 1.2 TB monthly cap evaporates fast: two people in videoconference-heavy jobs burn 400 GB, kids streaming school video add 300 GB, then Netflix and gaming eat the rest by week three. The throttling after cap is far worse than a slow connecal—it's unpredictable. One day you're fine, the next day your video call turns into a slide show. That's not a speed problem; it's a volume ceiling masquerading as a performance issue.

Most speed audits ignore data caps entirely. They measure a moment, not a month.

Look at your contract's fine print: does it say 'deprioritized after 50 GB'? That's a hard wall hidden behind a fast number. For low-income familie, overshoot fees or throttled speeds create a chilling effect—kids ration their usage, skip interactive assignments, and turn off cameras to 'save data.' The real access divide isn't speed. It's the fear of hitting a limit you can't afford.

Your next step: collect three months of actual usage from a sample of users—not speed check logs, but cellular data counters and ISP usage trackers. Compare that to the outline cap. If 40% of your community hits 80% of their cap before day 20, you don't orders more bandwidth. You need a different plan structure, or a community hotspot with uncapped hours for critical tasks.

Next Steps: From Audit to Action Without the Speed Hype

I once watched a town council celebrate '100 Mbps to every home' — then discovered the only library had closed its laptop-lending program because the new fiber ran past the door but no one could afford the monthly bill. Speed maps are a marvelous political prop. They let officials point at a color-coded grid and declare victory while familie still tether to a lone phone in the parking lot of a shuttered K‑Mart. The fix is unglamorous: publish a parity report that layers actual adoption data — not just availability — over the same geography. Show how many households subscribe at the lowest tier. Show how many devices the median household can more actual put online simultaneously. That hurts. Because when a speed map glows green but the parity report stays red, the pretense collapses.

— site technician, rural broadband co‑op

Invest in device libraries and digital navigators, not just bandwidth

Speed is seductive. A fat pipe looks like progress. But what usually breaks first is the device — a five-year-old Chromebook that can't hold a charge, a family sharing one phone for three remote learners. I've watched a school district spend $2 million on fiber upgrades while the actual homework gap widened because kids had no reliable screen. Wrong order. Redirect a fraction of that budget into a device library: loaner laptops, hotspots that aren't locked to a one-off provider, and digital navigators — real humans who help families set up Wi‑Fi filters, update drivers, or find the subsidy application buried three clicks deep on an ISP's site. The trade-off stings: you delay a speed upgrade that would score a grant milestone. But the seam blows out on access, not throughput. One navigator can keep thirty households online; a speed bump only matters if someone can actually log on to use it.

pull that ISPs publish real-world latency under load, not just peak speeds

Speed-test apps are liars. They measure a single clean connecing to a server ten miles away, not the evening scrum when four devices stream, two kids Zoom, and a VPN crawls. I've seen a '50 Mbps' connection fall apart at 7 p.m. because bufferbloat turned latency into a parking lot. The fix is a simple, brutal demand: force ISPs to publish latency under load — the round-trip time when the line is saturated — alongside the usual advertised speed. Most won't. That's the point. A number that jumps from 15 ms to 900 ms tells you more about real access than any 'up to' boast. One community I worked with made this a condition for a municipal franchise renewal; the ISP quietly offered a cheaper tier with traffic shaping instead of the promised symmetric gig. The parity report caught it. And that's where the work shifts from audit to pressure — concrete, public, impossible to spin.